In this study, we analyze the hysteresis behavior of robotic joints equipped with harmonic drivers under various speeds and loads, focusing on a range of hysteresis models including the Bouc–Wen (BW), Fractional Order Bouc–Wen (FOBW), and their integration with the Krasnoselskii–Pokrovskii (KP) model. This integration, which forms the KP-FOBW model, is a novel aspect of our research, offering an enhanced understanding of non-linear and hysteresis behaviors in robotic joints. We employ the Monte Carlo method, particularly focusing on the Stochastic Gradient Hamiltonian Monte Carlo (SGHMC), for precise parameter identification. Our results demonstrate that the FOBW model, especially when combined with the KP model, provides a more accurate representation of the hysteresis curves under varying operating conditions. The KP-FOBW combination emerges as a powerful tool to predict the output torque in robotic joints. This study contributes to a deeper understanding of hysteresis in joints with harmonic drivers, showcasing its potential for complex dynamic modeling in robotics.
18 March 2019 in Industrial Robot: the international journal of robotics research and application
Purpose: Remote handling (RH) manipulators have been widely studied for maintenance tasks in fusion reactors. Those tasks always require heavy load, high accuracy and large work space for manipulators. Traditionally, the maintenance of fusion devices always depends on manual RH. With the development of calculating ability, the intelligent automatic maintenance makes it possible for a fusion device instead of the previous manual operation. As the flexibility of arm and the deformation of manipulator will cause problems, which are mainly inaccuracy and lower efficiency. This paper aims to study an effective way to promote the arm behavior to solve these problems. Design/methodology/approach: By making use of the experimental advanced superconducting tokamak articulated maintenance arm as a platform, a series of experiments is designed to measure errors of kinematics and to collect the database of the flexible arm. Through studying the data and the arm structure, recurrent neural network (RNN) method was adopted to estimate the deformation of flexible arm and eventually compensate deformation in robot control to achieve higher accuracy. Findings: By means of delicate RNN modeling, errors of kinematics have been reduced to a smaller order than the RH mode. This intelligent maintenance method will also reduce complexity of operations in maintenance. Originality/value: This paper presents the use of an artificial intelligent algorithm to solve a nonlinear deformation problem of the flexible arm. The results demonstrate that it is efficient in dealing with this problem in fusion application. The RNN’s successful application has also shown that intelligent algorithms can be widely applied in fusion maintenance.
This paper proposes a design method for a miniature air–land amphibious inspection drone (AAID) to be used in the latest compact fusion reactor discharge gap observation mission. Utilizing the amphibious function, the AAID realizes the function of crawling transportation in the narrow maintenance channel and flying observation inside the fusion reactor. To realize miniaturization, the mobile platform adopts the bionic cockroach wheel-legged system to improve the obstacle-crossing ability. The flight platform adopts an integrated rotor structure with frame and control to reduce the overall weight of the AAID. Based on the AAID dynamic model and the optimal control method, the control strategies under flight mode, hover mode and fly–crawl transition are designed, respectively. Finally, the prototype of the AAID is established, and the crawling, hovering, and fly–crawling transition control experiments are carried out, respectively. The test results show that the maximum crawling inclination of the AAID is more than 20°. The roll angle, pitch angle, and yaw angle deviation of the AAID during hovering are all less than 2°. The landing success rate of the AAID during the fly–crawl transition phase also exceeded 77%, proving the effectiveness of the structural design and dynamic control strategy.
Although deep neural network (DNN)-based robot grasping has come a long way, the uncertainty of predicted results has prevented DNN-based approaches from meeting the stringent requirements of some industrial scenarios. To prevent these uncertainties from affecting the behavior of the robot, we break down the whole process into instance segmentation, clustering and planar extraction, which means we add some traditional approaches between the output of the instance segmentation network and the final control decision. We have experimented with challenging environments, and the results show that our approach can cope well with the challenging environment and achieve more stable and superior results than end-to-end grasping networks.
12 September 2023 in Fusion Engineering and Design
Robotic pick-and-place represents a nascent but swiftly evolving field in automation research. Most existing research relies on three-dimensional (3D) observations obtained directly from the 3D sensor or recovered by the two-dimensional (2D) camera from multiple perspectives. In this paper, we introduce an end-to-end pick-and-place neural network that solely leverages simple yet readily accessible data, namely, monocular camera and forward kinematics, for fusion applications. Additionally, our approach relies on the deep reinforcement learning (DRL) algorithm to facilitate robots in comprehending and completing tasks. The entire process is data-driven, devoid of any artificially designed task sessions, which imbues our approach with enhanced flexibility and versatility. The proposed method exhibits excellent performance in our experiment.